No label defined (Q4945433)

This article announces a beta version of the solver \(Sdpsol\) for linear matrix inequality programming. The constraints of the problem are linear matrix inequalities, the objective function may depend on the matrix as well. The authors examine the so called max-det problem formulated as \[ \begin{aligned} \min c^Tx+\sum_{i=1}^K & \log(\det G^{(i)}(x))^{-1},\\ G^{(i)}(x)>0,&\quad i=1,\dots ,K,\\ F^{(i)}(x)>0,&\quad i=1,\dots ,K,\\ Ax=&b, \end{aligned} \] where \(\mathbf x\) is \(m\)-vector, \( G^{(i)}(x)\) is a matrix function, \( G^{(i)}(x)=G^{(i)}_0+x_1G^{(i)}_1+\dots +x_mG^{(i)}_m\), the \(G^{(i)}_j\) are symmetric matrices. \( F^{(i)}(x)\) is defined similarly. A special case of the problem is the semidefinite program. The solver \( Sdpsol\) supports a language of description and transformation of data similar to Matlab one's. Examples of notation in \( Sdpsol\) and manipulation with matrix constraints are demonstrated. The authors apply \( Sdpsol\) to a real mechanics problem of mass-spring damper analysis and experimental design. The analysis of the damper model is reduced to the stability of a system of linear differential equations and then to a Lyapunov matrix inequality. The optimal design of the experiment is directly modeled as max-det problem. For the entire collection see [Zbl 0932.00034].